Nice, I have noticed the same thing when working with canvas - any time you can get the pixel level manipulation to happen in native objects there are big performance gains to be had.

I am interested on the general idea behind your new import an existing site feature. I have played around with this a little bit before, and realize that there is no way to directly render HTML onto a canvas (unless of course you have add on permissions in firefox, like this: https://developer.mozilla.org/en/drawing_graphics_with_canva...). Then of course there is something like this: http://www.cuppadev.co.uk/projects/a-browser-within-a-browse... which seems a little incomplete.

I assume you build up the elements array from the elements in the DOM, and manually generate a canvas for each one? That seems like it would be quite tedious, and handling inherited styles, etc could be a pain. Also, once you start moving elements around, they may not really have a place in the HTML anymore, so if I want to share the mockup, is it just an image, or does it become absolutely positioned HTML elements?

Great job on this by the way, I'm impressed with how well it is working.